From:	sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id:	<9701290239.AA13583@clio.rice.edu>
Subject:	Re: CWSDPMI memory capacity
To:	fesenko AT po DOT pacific DOT net DOT sg (Victor Fesenko)
Date:	Tue, 28 Jan 1997 20:39:22 -0600 (CST)
Cc:	djgpp AT delorie DOT com
In-Reply-To:	<19970129011336.AAA28922@fesenko.pacific.net.sg> from "Victor Fesenko" at Jan 29, 97 09:13:38 am

> I'm sorry for misunderstanding, what I really am interested in is the
> ability to use more than 128Mb of RAM without any paging to hard drive.
> I know that there is NEVER enough RAM for a programmer, but anyway ;-)
> My PC can support up to 512Mb of RAM and I beleive this is not a
> limit. And I also have a memory manager which supports 256Mb of RAM

I just looked at valloc.c - and except for the size of the dimension
for map[], it looks like 256Mb of physical memory should work also.
int15 and XMS can't show more than 64Mb, but VCPI can - and if it does
show more than 128Mb you can trigger a bug in valloc (overwrite the map
array).  I really should have malloc()ed this array.

With VCPI and more than 256Mb, you have to go to long ints to handle the
page number.  This would not be too hard to do, but it would be prone
to bugs.  Right now, if you had more than 256Mb, you would get the
modulo 256Mb size.

So, to increase the physical RAM handling to 256Mb, increase the
dimension of map to 4096; also increase the same constant in the memset
below.  Have fun.

- Raw text -